“Large language models are built to hallucinate.”

Where do you get your daily news – from the radio in the bathroom in the morning? At the breakfast table with a daily newspaper? Or do you scroll through your smartphone feed and click away as soon as you hit a paywall? The way we consume information has fundamentally changed in recent years. The circulation of daily and Sunday newspapers has more than halved since 1995 in Germany and beyond, and sales of books, newspapers and magazines are also declining.

At the international level, the picture doesn't look any better. The media industry is in a state of upheaval. It is increasingly difficult for quality journalism to rise above the noise. Competition, time and cost pressures in the media sector have been increasing for years. Declining advertising revenue in the print sector – which has not yet been offset by additional revenue in the booming online sector – is making life difficult for many media companies. These changes have an impact on the labour market and employment conditions, writes the German Federal Labour Office in its most recent look at the industry.

Can a media company still offer high-quality and creative content? Or will none of my morning reading be written by humans anyway?

Hello, who is writing?

Back to the kitchen table for the news of the day. So there I am, coffee in hand, scrolling through the feed, not really knowing if I can trust what I’m reading.

Can I trust what I’m reading?

Recently, several high-profile cases have revealed that, despite the difficulty in finding concrete evidence, some of their content was created by artificial intelligence (AI). Consumers felt betrayed and in times of “fake news” and general uncertainty, this could further undermine the credibility of media content.

Will these employees of the future require a new skill set in addition to a new mindset? In other words: are completely new skills required? “I’m not sure if you even need any additional skills,” Markert says. “The only thing you need is curiosity.”

With the aim of inspiring employees to take this journey, BurdaForward held its first AI Week last year, a large-scale internal training week where participants learned about quantum computing, cyber robotics, and the use of generative AI. Hundreds of employees and well-known speakers, companies and technology providers took part. Additionally, a biweekly AI Tuesday – which sees around a hundred colleagues exchanging knowledge, testing tools and sharing current developments and perspectives – ensures that everyone in the company stays on the ball. For young people in the company's own journalism school, artificial intelligence is already a natural part of the curriculum and practice.

The only thing you need is curiosity. Oliver Markert

If the synthesis succeeds

The use of AI, no matter how disruptive it is perceived today, is the next logical step in the interests of the users – comparable to the invention of the computer mouse or the cell phone camera, Markert is convinced: “Soon you won’t even notice the use of AI because it will be so seamlessly integrated,” he says.  “This is already happening in devices today – there are so many parts of them that we use without question. Like cell phone cameras, for example: they run so many scripts in the background, yet we simply press a button, and the device takes a picture. Not too long ago, taking a picture involved setting the aperture and exposure time, then waiting two weeks for your photos to be developed. That's all gone now. It will be the same here too: AI will be in the centre, but run in the background.  As an employer, our job is to design the best work software that our teams enjoy working with and produces results.”

“The interaction of human creativity and digital tools can set off brilliant fireworks.” This is how Miriam Meckel and Léa Steinacker sum it up in their book “Everything everywhere at once: How artificial intelligence is changing our world and what we can gain”. Meckel, a Professor of Communication Management at the University of St. Gallen, and Steinacker, a PhD in Social Sciences, showcase the opportunities that can arise from a successful synthesis of human potential and technological innovation.

Duty of care against fakes

But what if AI is used for conscious manipulation? Just recently, Google blew the whistle on its chatbot Gemini, which had put a person of colour in a Nazi uniform when asked to show a German soldier from 1943. If you follow a recently published representative study by the Bertelsmann Foundation, its survey results show “a growing distrust of the media and politics and, at the same time, uncertainty.” Almost a third of respondents have little trust in the media. Awareness of the risks of intentional misinformation for democracy has increased in large parts of the population, according to the study.

Markert doesn’t think this is new. “If we look at the history of journalism, there have always been attempts to manipulate journalists. That's why we have smart people sitting there who are not easily hoodwinked. We are constantly sensitizing our people to the fact that no photo, text, video or voice can be trusted. We are already working with the big players to properly assign the source of images from image-generative AI, like Adobe Firefly, for example, and to quickly recognize any changes to the image. Because we know how these technologies work, we also know the pitfalls and are preparing ourselves in the best possible way to avoid stumbling across a counterfeit.”

Tool or generator?

But AI is only one tool among many in the process, with humans holding the reins. “The journalistic duty of care includes clear source information; that hasn’t changed,” Markert explains. “For example, if an illustration comes out of a machine, our convention is that it says Dall-E or Midjourney. In our company there is no text content created purely with AI, but always people who check both the input and the output.”

What rules does the new game need?

A ruling by the Press Council, which reprimanded a cooking magazine for using AI-generated images in December 2023, says that binding and accepted rules of the game have yet to be established. That same month, the New York Times spectacularly sued openAI for copyright infringement. For years, Google boss Sundar Pichai has been calling for international rules for artificial intelligence.

Hubert Burda Media has developed a method of dealing with AI. The company is committed to credibility and has created an AI statement that proves that facts have been properly researched and sources are reputable.

What does the priceless asset of credibility mean in the race for users’ attention? Will other providers be faster if they don't care about the veracity of a message? Markert is unconcerned. “If the BBC reports that the Queen is dead, the Queen is dead. There are absolutely reliable sources, for example from official institutions, that you can simply trust. And, of course, we have people in our editorial team who call, ask questions and check whether the story is watertight. We definitely take the time. These are classic journalistic research qualities and a basic human job for us, even in times of AI.”

The general counsel and deputy managing director of the German Association of Journalists, Hanna Möllers, recently said something similar. “AI cannot perform the core tasks of journalism: Analyse political connections, identify problems, mediate between politics and the population. Everything that concerns the watchdog tasks of journalism in a democracy.” Reporters Without Borders, for example, has launched an initiative based on ISO standards to ensure the credibility of journalism. Microsoft uses the standards in its algorithms to ensure that independent journalism is given more prominence, says democracy advocate Michael Bak.

Hallucination needs safeguards

Has anything ever gone wrong at BurdaForward? “Of course, it’s very important to tell employees to make sure the technology doesn’t make a mistake,” says Markert. “We know that our system invented protagonists and quotes that weren’t even there in the original. It is a design task: what went into the machine and what comes out again?

You have to remain sceptical.Oliver Markert

We tell our colleagues very clearly that the large language models are built to hallucinate. However, our task is to set safeguards through our technical specifications to ensure that they do not do this, if possible. But we will never be able to achieve this 100 percent. That's why we need people. You have to remain sceptical.

We tell our colleagues: ‘We can always speed up the first 80 - 90 percent of the way, but for the last 10 percent we need you and your full concentration’. I’m convinced that this is an opportunity for us as a media company to differentiate ourselves from these other platforms.”

From creative writer to supervising fantasising AI models, is this the current career path for journalists?

Human is simply a guarantee of quality.Oliver Markert

The needs of the users

“Our former boss always said: ‘At its core, digitalization is not a technological revolution, but rather the radical placing of people at the center’. We are very lucky that we are a technology-loving company – this has always opened up opportunities for us to see developments at an early stage, to embrace them, try them out, incorporate them and thus create value,” summarises Markert. “Human is simply a guarantee of quality, something that won’t – and shouldn’t – go away in the foreseeable future.”

And perhaps this is the way to make an entire industry viable.

The interview with Oliver Markert was conducted by Maike Tippmann. 04/2024